[clang] [AMDGPU] Mark int64_t and intmax_t as LP64 (PR #214865)

[Yaxun Liu via cfe-commits]

yxsamliu wrote:

Could we also add a direct test for the predefined types?

The current test updates check this through C++ mangling / builtin parameter types, but the actual behavior changed here is the target definition of `__INT64_TYPE__` and `__INTMAX_TYPE__`, which then affects `stdint.h`. A small `-E -dM` or `static_assert` test for standalone `amdgcn-amd-amdhsa` would make the intended behavior clearer.


https://github.com/llvm/llvm-project/pull/214865